I'd especially like to have your advice (do not hesitate to be harsh :) on the new publisher module loading system.
It does not uses apache.import_module, and fixes a lot of nasty bugs. However, the new system means that standard Python modules and published pages are not the same things anymore. Well, published pages *are* Python modules, but they are not registered into sys.modules. This breaks some of the import keyword semantics : importing a published page from another published page cannot be done using "import", but through the new publisher.get_page(req,path) function. "import" can of course still be used to import anything that can be found on the PYTHONPATH. I really think this is the way to go, and both mpservlets and Vampire are also separating standard Python modules from dynamically loaded modules. But if someone feels this is wrong, please tell me ! Regards, Nicolas